CarrieLab

Interesting. I wonder how far we can push the "AI-generated UI" pattern with today's models. Is GPT 3.5 good enough for or will we need GPT 4, and if so, will it be fast enough (I assume yes, eventually)?

I'm a PM at a human data company (https://www.surgehq.ai) that helps the large language model companies ensure their models are safe (we're the “clever prompt engineers” who helped Redwood assess their model performance).

We actually just published a blog today that includes our perspective on building “AI red teams” and best practices for AI alignment/safety: https://www.surgehq.ai/blog/ai-red-teams-for-adversarial-tra...

Agreed

Operational overhead and increased liability for lawsuits, I imagine.

Think this just shows you truly important SEO is (to media companies). It's probably easily worth the trouble for them if there are SEO gains.

ty!

Ah thank you!

How do you use it to bookmark?

Definitely not a surprising result. Though I don't think the article is claiming it is a surprise.

I appreciate that he's drawing clear lines (aside from the generically "severe" consequences promised in response to Russia using nukes, which seems like sensible strategic ambiguity). Have to wonder what the game plan is if Russia does indeed use nuclear weapons. All options seems terrible.

For ipad, reading only (no social media, slack, etc).

Breitbart.com on HN! That's new.

Oh I think your high school pop music theory is spot on. I grew up in the emo era, can actively laugh at the music/fashion now... and still love it more than anything :)

Have you seen this post / these examples? It touches on some what of you are talking about in terms of style transfer.

https://www.lesswrong.com/posts/r99tazGiLgzqFX7ka/playing-wi...

> I don't think you'll have many takers here suggesting that things were magically better 40 years ago.

Ha, fair point. I must not realize how old I am, because I was attempting to reference the music of the 1960s and 70s, not 1982, which I agree is not many people's idea of the golden year for music ("Come On Eileen" notwithstanding).

> Sophisticated tools are a bit of a trap. People tend to create in ways that their tools make easier.

No doubt. Ableton, logic, and protools have drastically altered the norms of what modern music is "supposed" to sound like (ie tuned vocals, quantized drums etc). I do wonder what the next generation of music tech will bring.

My intuition is that humans will continue to make art that takes advantage of technological advances, just like they always have.

The modern process of producing music would basically be unrecognizable to anyone 40 years ago — it's completely intertwined with technology, and far more automated. Yet music is as important as ever, and amazing music is being made (will politely side-step the pitfall of debating whether music was better 40 years ago!)

So I'm excited to see how visual artists incorporate tools like Dall-E into their artistic process.

DALL-E 2 is a true "holy shit" moment for me. It's actually hard to believe it's real.

I'm on the team at Surge AI (data labeling platform + workforce), so this article hits home. We started Surge AI precisely because our team has always been against the adversarial, penny labor design of crowdwork systems: systems dependent on multi-annotator consensus lead to poor quality data (and ignore the inherent subjectivity in the rich, language-based tasks I love), they lead to suboptimal outcomes and treatment for all parties, and you get get what you pay for!

For example:

1. Most data labeling systems don't allow you to communicate meaningfully with your workforce. In contrast, we prize two-way communication; your data labelers are the ones going through tens of thousands examples, so they often have amazing feedback for how to improve your data and design your tasks better. And of course, you often have questions for them as well.

2. Context matters. I can't label Spanish hate speech; someone from Mexico City often can't label Madrid slang either.

3. The majority vote isn't always the best one. Real-world data is often personalized and subjective; your opinion on a funny or angry story may not match mine, and that's okay. Our training sets and AI models should reflect that. We just wrote a blog post on the subtle nuances when considering majority votes and inter-rater reliability metrics: https://www.surgehq.ai/blog/the-pitfalls-of-inter-rater-reli...

4. Curating annotator pools. Our product is designed around helping you build custom labeling teams that you trust, who learn the nuances of your domain and stay with you over time.

Exact same experience for me (though I probably would have bought it either way). My sense is that the reddit vitriol is from a vocal but small minority.

Huge fan of infinite's multiplayer. They knocked it out of the park in my book. Need splitscreen support to seal the deal though — Splitscreen multiplayer is the ultimate halo experience, in my mind.

"If Spotify intends to dominate the industry, then the industry is in trouble."

Funny to suggest that Spotify may NOT intend to dominate the industry? Surely that's their goal.

I often see subtle misuses of interrater reliability metrics.

For example, imagine you're running a Search Relevance task, where search raters label query/result pairs on a 5-point scale: Very Relevant (+2), Slightly Relevant (+1), Okay (0), Slightly Irrelevant (-1), Very Irrelevant (-2).

Marking "Very Relevant" vs. "Slightly Relevant" isn't a big difference, but "Very Relevant" vs. "Very Irrelevant" is. However, most IRR calculations don't take this kind of ordering into account, so it gets ignored!

Cohen's kappa is a rather simplistic and flawed metric, but a good starting point to understanding interrater reliability metrics.

Always a must read